When analyzing data in the vast majority of knowledge domains, it is common to encounter a high number of zeros. This is no different when working with data science in agriculture. When dealing with proportional data with a zero inflation, a useful approach is to model the problem with zero-inflated beta regression (ZIBe). This allows for a statistically correct (or at least reasonable) approach and enables the counting of zeros in the database. In this study, we analyzed a dataset gathered from a field experiment, aiming to ascertain the average proportion of citrus canker present on orange plant leaves. This was done in relation to the genotypes of four different rootstocks used in the experiment. The experiment combined the genetics of four rootstocks (lower part of the plant) with nine types of orange varieties in the canopy (upper part of the plant). Modeling provided information regarding the estimation of the expected mean value through modeling with Bayesian zero-inflated beta regression. This made it possible to assess the average incidence for a given plant based on its genotype and rootstock combination, allowing for the estimation of the expected value for the observed combination. Upon concluding the modeling, it was observed that the Orange Caipira rootstock genotype appeared to be more resistant to the disease, while the Lemon Cravo rootstock genotype was classified as the most susceptible. The rootstock genotypes Ol ímpia, Arapongas, Ipigu á IAC, and EEL had equal chances of developing the disease. Supplementary materials accompanying this paper appear on-line.